Problem solving and decision making

Superintendents make dozens of decisions a day. Most are small and routine: which crew goes where, whether a delivery can come at 10 instead of 8, how to route a hose around a stair. Some are large and uncertain: whether to pour a slab with rain forecast, how to recover a two-week delay, what to do when a beam arrives the wrong length. And many decisions begin as problems, something has gone wrong and needs fixing, without making things worse. This lesson offers a structured way to solve problems, tools for finding root causes, methods for making decisions under uncertainty, the biases that lead good people to bad decisions and guidance on when to decide and when to escalate.

BUS-130 introduces decision making in organizations; CON-410 covers the A3 problem-solving report and continuous improvement in lean construction; CON-345 covers quality nonconformances; CON-420 covers what happens when problems become claims. This lesson is about the supervisor's everyday problem solving in the field.

A structured approach to site problems

Under pressure, people jump from symptom to solution. The concrete is honeycombed, so patch it; the drywall crew is slow, so add workers; the inspection failed, so fix the item and rebook. Sometimes that is enough. But problems that come back, cost a lot or involve safety deserve a more structured approach. A simple six-step cycle works for most site problems.

Defining the problem

A well-defined problem is half solved. Charles Kepner and Benjamin Tregoe, whose methods are still taught widely, recommended describing a problem by what it is and what it is not: where it occurs and where it does not, when it started and when it does not happen, how big it is and how big it could be. The contrasts point to causes. If drywall screw pops appear on level 3 but not levels 2 or 4, in units on the north side but not the south, and only in ceilings, something specific to those ceilings on level 3 is different, perhaps a different framing crew, a different joist supplier or a period of high humidity when the board was hung.

It also helps to state the problem in terms of its impact on the project's goals: 'level 3 north ceilings show screw pops in about 30 % of units, which will fail the owner's finish inspection and delay turnover of 14 units' is a problem statement that tells everyone why it matters and how big it is.

Contain first

For safety and quality problems, the first action is containment: stop the problem from growing while the cause is found. If a guardrail post is found loose on one floor, the immediate action is to secure the area and check other posts installed by the same crew, before investigating why. If a batch of concrete cylinders breaks low, the immediate action is to identify where that concrete was placed and whether further loading should wait for engineering advice. Containment is not the fix; it buys time to find the real fix without adding damage.

Finding root causes

The immediate cause of a problem is rarely the whole story. A pour is delayed because the pump truck was late; the pump truck was late because it was booked for the wrong day; it was booked for the wrong day because the pour date changed and the change was communicated by text to someone who was on holiday. Fixing only the immediate cause, by shouting at the pump company, does nothing to stop the next missed booking. The root cause is the weak communication of schedule changes.

The five whys, a technique associated with Toyota and Taiichi Ohno, asks 'why?' repeatedly, usually about five times, until it reaches a cause that can be fixed in the system. It is simple and quick, but it has a weakness: it tends to follow a single chain and can stop at a person ('because Jim forgot'). Good practice is to keep asking why Jim forgot, which usually reveals a system weakness, and to consider several branches.

The fishbone or cause-and-effect diagram, developed by Kaoru Ishikawa, helps teams think more broadly. The problem is written at the head of the fish and the main bones represent categories of cause. Manufacturing uses the 6Ms (manpower, methods, materials, machines, measurement and mother nature); for construction a useful set is people, methods, materials, equipment, information and environment. The team brainstorms possible causes under each heading, then tests the most likely ones with evidence.

Making decisions

Many site decisions are programmed: they follow established rules or company procedures, such as how to respond to a near miss or when a hot work permit is needed. Others are non-programmed: new, unstructured situations that need judgment. Non-programmed decisions benefit from a few disciplines. Generate more than one option before choosing; the first idea is rarely the best. Be clear about the criteria that matter, such as safety, cost, schedule, quality and relationships. Involve the people who will carry out the decision or who know the most. Consider what could go wrong. And decide at the right speed.

A useful distinction, popularized by Jeff Bezos, is between reversible decisions, which can be undone cheaply if they turn out wrong, and irreversible ones, which cannot. Reversible decisions, such as moving a laydown area or trying a new crew arrangement for a week, should be made quickly by the person closest to the work. Irreversible decisions, such as cutting a post-tensioned slab, pouring concrete or removing shoring, deserve more care, more information and often a second opinion or engineering approval. Many site delays come from treating reversible decisions as if they were irreversible; many site disasters come from treating irreversible decisions as if they were reversible.

Comparing options with a decision matrix

When options differ on several criteria, a weighted decision matrix makes the trade-offs visible. List the criteria, give each a weight reflecting its importance, score each option against each criterion and calculate weighted totals. The result is not an answer to be followed blindly, but it forces the team to state what matters and why, and it often reveals that the favorite option is weaker than it seemed.

On Larkspur Commons, the wood frame is running two weeks behind and Dana is considering three recovery options: a second framing crew from another subcontractor on the upper floors, a Saturday shift for the existing crew for six weeks, or resequencing so that mechanical and electrical rough-in follows framing floor by floor more closely.

# Weighted decision matrix for three schedule recovery options (scores 1-5, higher is better)
criteria = {"safety": 0.30, "schedule recovery": 0.25, "cost": 0.20, "quality risk": 0.15, "relationships": 0.10}
options = {
    "Second framing crew":     {"safety": 3, "schedule recovery": 5, "cost": 2, "quality risk": 3, "relationships": 2},
    "Saturday shift (6 weeks)": {"safety": 2, "schedule recovery": 3, "cost": 3, "quality risk": 3, "relationships": 3},
    "Tighter resequencing":    {"safety": 4, "schedule recovery": 3, "cost": 5, "quality risk": 4, "relationships": 4},
}
assert abs(sum(criteria.values()) - 1) < 1e-9
results = {}
for name, sc in options.items():
    results[name] = sum(criteria[c] * sc[c] for c in criteria)
for name, total in sorted(results.items(), key=lambda kv: -kv[1]):
    print(f"{name:26s} weighted score {total:.2f}")
# Sensitivity: does the ranking change if schedule recovery matters much more?
alt = {"safety": 0.25, "schedule recovery": 0.45, "cost": 0.10, "quality risk": 0.10, "relationships": 0.10}
print("If schedule recovery is weighted 45 %:")
for name, sc in options.items():
    print(f"  {name:26s} {sum(alt[c] * sc[c] for c in alt):.2f}")

Tighter resequencing       weighted score 3.95
Second framing crew        weighted score 3.20
Saturday shift (6 weeks)   weighted score 2.70
If schedule recovery is weighted 45 %:
  Second framing crew        3.70
  Saturday shift (6 weeks)   2.75
  Tighter resequencing       3.65

With the original weights, tighter resequencing scores best: it is safer, cheaper and kinder to relationships, though it recovers less time than a second crew. The sensitivity check shows that if recovering every day of schedule mattered far more than cost, the second framing crew would edge narrowly ahead, 3.70 against 3.65, so the ranking depends on how much the owner values each day recovered. In practice Dana might combine options: resequence now, and line up a second crew in case the resequencing does not recover enough within three weeks. The matrix did not make the decision, but it focused the discussion with Priya on what mattered.

Deciding under uncertainty

Many site decisions involve uncertainty: weather, inspection outcomes, delivery reliability. One way to reason about them is expected value: multiply each possible outcome's cost by its probability and add them up. Consider a decision to pour the level 2 slab tomorrow with a 30 % chance of heavy rain. Pouring and getting lucky costs nothing extra; pouring in heavy rain might cost $40,000 in surface repairs and protection; postponing costs about $9,000 in crew time, pump rebooking and schedule. The expected cost of pouring is \(0.30 \times 40{,}000 = 12{,}000\), higher than the $9,000 cost of postponing, so on expected value alone the pour should be postponed, unless the forecast improves or better protection can reduce the damage.

Expected value is a useful discipline, but it has limits. Probabilities are often guesses, and some outcomes are so severe that their expected cost understates them: a small chance of a structural failure or a fatality is never acceptable just because its probability-weighted cost is small. For safety-critical decisions, the right question is not 'what is the expected cost?' but 'can we make this safe?'. Herbert Simon's idea of satisficing is also relevant: in real conditions people rarely find the optimal choice, and a good-enough decision made in time often beats a perfect one made too late.

Biases that trap good supervisors

Psychologists such as Daniel Kahneman and Amos Tversky have documented systematic biases in human judgment. Several are especially dangerous on construction sites.

Normalization of deviance, a term coined by sociologist Diane Vaughan in her study of the Challenger space shuttle disaster, deserves special attention. It describes how people gradually come to accept a departure from a standard because nothing bad has happened yet. A crew that skips one shoring check under time pressure, gets away with it and skips it again has normalized the deviance. Each time feels safe because the last time was. Supervisors counter it by comparing current practice with the standard, not with recent habit, and by treating near misses as warnings rather than proof that the shortcut is safe.

Decide, involve or escalate

Supervisors also need to judge who should make a decision. Some decisions belong to the superintendent; some should involve the crew or foremen; some must be escalated. Decisions should be escalated when they exceed the supervisor's authority (contract scope, money beyond a limit, commitments to the owner), when they involve design (any change to structural elements, fire separations, building envelope or anything the engineer or architect of record controls, which goes through an RFI or the consultant), when they involve engineered temporary works such as shoring and formwork designed by an engineer (CON-325), when they involve legal or regulatory questions, or when the consequences are severe and the supervisor is unsure.

Escalating is not a sign of weakness. Superintendents get into serious trouble not by asking too many questions but by making design or contract decisions they had no authority to make: telling a framer to drill through a beam, approving a substitute product without review, agreeing with an owner's representative to a change in finishes. The good habit is to escalate with a recommendation: 'here is the problem, here are the options, here is what I recommend and why'.

Finally, record important decisions: what was decided, by whom, on what information, and why. A note in the daily report or a short email takes two minutes and can save weeks if the decision is questioned later (CON-375 and CON-420). After significant problems, a short after-action review, asking what was supposed to happen, what actually happened, why there was a difference and what we will do next time, turns experience into learning.